Post:
Hi DEV friends 👋,
Over the past month I’ve been tinkering with an idea: Ml Forge , a desktop-first IDE for Vision AI.
The motivation is simple: most Vision AI workflows feel fragmented. You jump between scripts, configs, dataset tools, training dashboards, and export utilities. My goal is to unify that into one place — with a no-code but still controllable approach.
Here’s what I’ve got so far:
Dataset manager (versioning, splits, validation)
Annotation studio
Training engine with live GPU metrics
Reproducible runs (dataset → config → model)
Logs & metrics
Inference & benchmarking
Export to ONNX, TensorRT, CoreML, TFLite, OpenVINO
Everything is UI-driven — no training scripts or deployment code.
I’d love to hear your perspective:
Does “no-code but controllable” resonate for Vision AI workflows?
What would make a tool like this useful in your day-to-day?
What blockers or concerns do you see?
I’m still early in development, so feedback from this community would be super valuable. Happy to dive into technical details or design trade-offs if anyone’s curious.
This tone works well on DEV because it feels like you’re sharing a journey and inviting collaboration, rather than promoting a product.
Would you like me to also add a short “storytelling hook” at the start (like how you struggled with fragmented workflows yourself) to make it more engaging for DEV readers?
refer the website to get info
Top comments (0)